How to automate email in ProShop ERP
Short answer
Monitor the shared inboxes where RFQs, customer POs and vendor confirmations arrive, extract the structured fields with a model, and create draft quotes, customer POs or PO updates in ProShop through its API. A person reviews anything below a confidence threshold.
In a typical ProShop shop, work enters by email. An RFQ arrives with a PDF drawing, a customer PO comes as an attachment with its own layout, a vendor replies to a PO with a new ship date in a sentence. Somebody retypes each of these into ProShop, often at the end of the day.
That retyping is slow and error-prone, but it is also the most automatable part of the quote-to-cash loop. The trick is to automate the capture and leave the commitment, the price and the promise date, with a human.
What teams use it for
RFQ intake to draft quote
An RFQ email creates a draft quote in ProShop with customer, part number, revision, quantities and requested date, with the drawing attached.
Customer PO matching
An incoming PO PDF is matched to the open quote, differences in price or quantity are flagged, and a draft customer PO is prepared.
Vendor acknowledgements
Replies to purchase orders are read for confirmed dates and prices, and exceptions are queued for the buyer.
Certs and packing slips by mail
Material certs and packing lists emailed by suppliers are attached to the right PO and receiving record.
Customer change requests
Emails asking for a revision change or date change are classified and routed to the work order owner with the work order referenced.
ProShop ERP integration surface
- ProShop API (GraphQL in current releases) can create or update quotes, customer POs, purchase orders and attach documents; confirm which objects accept writes on your plan.
- ProShop also generates its own emails (quotes, POs) which are sent from ProShop or your mail server; reply handling needs reply-to addresses that map back to the record.
- Inbound mail is read from Microsoft 365 through Graph, from Google Workspace through the Gmail API, or from IMAP, entirely outside ProShop.
- Customer and vendor master records in ProShop are the source for matching sender domains, names and part numbers.
- Attachments need to be stored with the record; confirm the file size and type limits for ProShop document attachments.
How it works
- 1
Connect the mailboxes
Use delegated or application access to shared inboxes such as sales@ and purchasing@. Process only messages in a defined folder or with a defined label.
- 2
Classify each message
Decide if it is an RFQ, customer PO, vendor acknowledgement, cert, change request or noise. Keep the class and a confidence score.
- 3
Extract fields
Pull customer, PO number, line items, part numbers, revisions, quantities, prices and dates from the body and attachments. Validate against ProShop master data.
- 4
Match to ProShop records
Look up the customer, existing parts and open quotes. If the part is new, queue it for engineering instead of creating a part record automatically.
- 5
Create drafts, not commitments
Write drafts or notes through the API and attach the original email and files, then route to a reviewer with the extraction shown next to the source.
- 6
Learn from corrections
Log reviewer edits per customer so recurring layouts and abbreviations are handled better over time.
Example flows
RFQ to draft quote
Trigger: An email with a drawing PDF arrives in the sales inbox.
- The classifier tags it as an RFQ.
- The extractor reads part number, revision, quantities and need-by date.
- The customer is matched by sender domain and the part by number.
- A draft quote is created in ProShop with the drawing attached.
- The estimator is notified to review pricing.
Result: Estimators start with a populated quote instead of an empty form.
Vendor date change
Trigger: A supplier replies to a PO with a new ship date.
- The reply is matched to the PO by number in the subject or thread.
- The new date is extracted and compared to the due date.
- If it slips past the work order need date, the buyer and planner are alerted.
Result: Schedule risk is caught the day the email arrives.
Keeping data on-prem
Email content carries customer names, pricing and drawings, which is often the exact material shops need to keep private. Run the extraction pipeline on your own servers, using an open-weight model served with vLLM or Ollama on local GPUs together with an OCR engine for attachments, and pull mail through your own mail server or an authenticated Graph connection. Only the structured result, not the raw mail, needs to reach ProShop through its API. For export-controlled drawings, keep them in your controlled store and attach a link or reference rather than copying them into a general-purpose system.
Security checklist
- Give the mailbox connector the least privilege possible, limited to specific shared mailboxes or folders.
- Never act on instructions found inside an email body, such as change this bank account. Treat email text as data, not commands.
- Require human review for new customers, new vendors, price changes and bank detail changes.
- Keep an audit trail linking each ProShop change to the source email and the reviewer.
Ways to get it done
Mail rules plus ProShop email or CSV import
Effort: Days of configuration.
Enough if one or two customers send consistent, structured POs.
Build an in-house pipeline with an LLM and OCR
Effort: Four to eight weeks for RFQ and PO intake.
Good for teams with a developer who can maintain extraction prompts and review screens.
Partner build with Netray
Effort: Pilot on one inbox in a few weeks.
Suits shops wanting private, on-prem extraction with a review queue and ProShop write-back.
Pitfalls
- !Auto-creating new part records from RFQs, which pollutes the part master.
- !Trusting extracted prices without comparing to the quote, so customer price errors slip through.
- !Processing mail threads without deduplication, creating several quotes from one RFQ and its follow-ups.
- !Treating prompt injection in an email body as harmless.
Frequently asked questions
Can ProShop read emails by itself?
ProShop sends emails, but turning inbound customer POs or RFQs into records generally needs an external layer reading the mailbox and writing through the ProShop API. Check your plan for API access and which objects can be created programmatically.
How accurate is automated PO extraction?
It varies by document quality and layout. Clean PDFs from repeat customers can be very reliable, scans and free-form emails less so. Use confidence thresholds, validate against master data and measure reviewer correction rates before widening scope.
Does the system auto-send quotes or accept POs?
It should not. Create drafts, show extraction beside the source, and have a person confirm price, date and terms. Automation handles the typing, people retain the commitment.
What about drawings attached to RFQs?
Store them with the draft quote or in your controlled file system, linked by reference. Reading dimensions from drawings is a separate and harder problem. Capture the metadata, part number and revision first and leave estimating to people.
Can this run without sending mail to a public AI service?
Yes. Use a locally hosted open-weight model and a local OCR component, with the mailbox read over your own connection. Only extracted fields move to ProShop, and original mail stays in your mail system.
Want Email automation working on your ProShop ERP?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment